"Sub-millisecond latency to major trading venues" is printed on half the VPS sales pages I've read this year. It's usually true in a narrow sense and nearly useless in the way most of us read it.
What the number usually describes is the network round trip between the hosting provider's data centre and a particular piece of infrastructure, often a well-known exchange or broker data centre in London, New York or Tokyo. If your broker's trade server sits in that same building, great. If it sits somewhere else, the advertised figure tells you very little.
What actually matters for my orders is the full chain: my click, the terminal, the VPS network, the broker's bridge, their pricing engine and back. Network ping is one link. On my own logs the network part is around 1 ms from a London VPS, yet the time from sending an order to seeing it confirmed in the terminal is typically 30 to 90 ms. Most of that delay isn't something hosting can fix.
So now I ignore the brochure and test ping to my broker's actual trade server, plus order confirmation times during 08:00 to 08:15, for a week.
A provider that gives me 2 ms every time beats one that gives me 0.4 ms most of the time and 40 ms when London opens.
What low latency hosting claims mean in practice
-
LondonScalper
- Posts: 965
- Joined: Sat Sep 05, 2026 7:54 am
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: What low latency hosting claims mean in practice
The 30 to 90 ms confirmation time is the number most people never look at, so thanks for putting real figures on it. In my logs the spread of that number matters more than the average.
I tested two providers for a week each at the London open. One averaged 48 ms to confirmation with a worst case of 70. The other averaged 31 ms but had six orders over 300 ms in the first five minutes after 08:00. I kept the slower one.
Another edge case worth mentioning is the terminal itself. An MT4 with dozens of charts and a couple of heavy indicators on a small VPS can add more delay than the network. When I cut my VPS terminal down to four charts with no custom indicators, confirmation time dropped noticeably.
I'd also check the time sync on the VPS. If the clock drifts, your logs lie to you.
When you measured 08:00 to 08:15, did you split market orders from pending orders? My pending fills behaved quite differently.
I tested two providers for a week each at the London open. One averaged 48 ms to confirmation with a worst case of 70. The other averaged 31 ms but had six orders over 300 ms in the first five minutes after 08:00. I kept the slower one.
Another edge case worth mentioning is the terminal itself. An MT4 with dozens of charts and a couple of heavy indicators on a small VPS can add more delay than the network. When I cut my VPS terminal down to four charts with no custom indicators, confirmation time dropped noticeably.
I'd also check the time sync on the VPS. If the clock drifts, your logs lie to you.
When you measured 08:00 to 08:15, did you split market orders from pending orders? My pending fills behaved quite differently.
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: What low latency hosting claims mean in practice
Something I'd add to the hosting checklist is the provider's maintenance window, because the best ping is useless during a reboot. Mine restarted for updates at 07:55 on a Tuesday, a quiet hour in the provider's own time zone and the worst possible one in mine.
The terminal came back up, but without the EA attached, because I had never set it to load on startup. I only noticed when a trade I expected to be managed was sitting there untouched at 08:20.
Now I ask support three things in writing before I pay. When is your maintenance window in UK time, do you give notice before reboots, and can I control Windows updates myself?
The answers sort providers quickly. One gave me a fixed Saturday slot with 48 hours notice, another said reboots happen when needed, which told me enough.
I also run a heartbeat check. If the EA sees no tick for two minutes during London hours it emails me, and a script on my laptop checks the VPS answers every five minutes.
The frozen terminal is the case that worries me most, because everything looks normal until you check. Has anyone found a cleaner way to catch a VPS that has quietly hung rather than gone fully offline?
The terminal came back up, but without the EA attached, because I had never set it to load on startup. I only noticed when a trade I expected to be managed was sitting there untouched at 08:20.
Now I ask support three things in writing before I pay. When is your maintenance window in UK time, do you give notice before reboots, and can I control Windows updates myself?
The answers sort providers quickly. One gave me a fixed Saturday slot with 48 hours notice, another said reboots happen when needed, which told me enough.
I also run a heartbeat check. If the EA sees no tick for two minutes during London hours it emails me, and a script on my laptop checks the VPS answers every five minutes.
The frozen terminal is the case that worries me most, because everything looks normal until you check. Has anyone found a cleaner way to catch a VPS that has quietly hung rather than gone fully offline?
-
Shadow Trader
- Posts: 5594
- Joined: Wed Sep 09, 2026 5:19 pm
Re: What low latency hosting claims mean in practice
Treating trading as a small business means listing costs that rarely appear in a backtest. Spread and commission are visible on each trade, yet platform fees, data subscriptions, charting tools, VPS hosting, news terminals, and evaluation fees also belong in the ledger.
For a scalper, data quality matters more than for slower styles. Tick-level history for testing, depth of market views, and low-latency hosting can each carry monthly charges. Some are worth it, others are comfort purchases. A useful test is whether the tool has changed a decision or measurably reduced execution cost in the past quarter.
A simple monthly statement helps: gross trading result, minus commissions and swaps, minus subscriptions and hosting, minus fees, equals net business result. Dividing fixed monthly costs by the average net profit per trade gives a break-even trade count, which reveals how much activity is needed before any real earnings arise. Smaller accounts feel this most, since fixed costs weigh heavily against modest absolute profits.
Reviewing subscriptions every few months, cancelling overlapping tools, and favoring free or broker-provided data when adequate keeps overhead low. The goal is not austerity but clarity, so that performance is judged after costs rather than before them.
For a scalper, data quality matters more than for slower styles. Tick-level history for testing, depth of market views, and low-latency hosting can each carry monthly charges. Some are worth it, others are comfort purchases. A useful test is whether the tool has changed a decision or measurably reduced execution cost in the past quarter.
A simple monthly statement helps: gross trading result, minus commissions and swaps, minus subscriptions and hosting, minus fees, equals net business result. Dividing fixed monthly costs by the average net profit per trade gives a break-even trade count, which reveals how much activity is needed before any real earnings arise. Smaller accounts feel this most, since fixed costs weigh heavily against modest absolute profits.
Reviewing subscriptions every few months, cancelling overlapping tools, and favoring free or broker-provided data when adequate keeps overhead low. The goal is not austerity but clarity, so that performance is judged after costs rather than before them.